Bat algorithm and neural network for monthly streamflow prediction

Streamflow prediction has a significance influence on improving water supply management and flood prevention. The applications of artificial intelligence (AI) have been proved to have better performance as compared to conventional statistical method in streamflow prediction. Therefore, this study proposed on the development of streamflow prediction model AI techniques namely Bat algorithm (BA) and backpropagation neural network (BPNN). BA is an optimization technique, which is to optimize BPNN in deciding optimum parameters and then improve the prediction accuracy. The study area chosen is Kuantan river and Kenau river, located in Kuantan, Malaysia. Two prediction models are proposed in this study which are BPNN and hybrid Bat-BPNN. Monthly historical rainfall data, antecedent river flow data and meteorology parameters data for two different rivers were used as the input to the proposed models. The performance of the proposed prediction models for Kuantan river and Kenau river are then being compared and evaluated in term of RMSE and R2. It is found that hybrid model, Bat-BPNN yields lower RMSE and provides higher R2 as compared to BPNN model at both Kuantan river and Kenau river. Therefore, it can be concluded that, proposed hybrid model yields better performances as compared to BPNN model for monthly streamflow prediction.